Sarvesh Garimella, ACME AtronOmatic dba MyRadar
Sarvesh Garimella, Chief Scientist and Chief Technology Officer, ACME AtronOmatic dba MyRadar
This paper will detail multiple advances made to use small satellites for hazard alerting with detail provided about wildfire detection and alerting from Low Earth Orbit (LEO).
Historically, satellite observations have been constrained by the spectral resolution of conventional multi-spectral sensors (e.g., GOES, MODIS). Over the last decade, Hyperspectral Imaging Spectroscopy, which is the measurement of surface-reflected solar radiation in a contiguous, narrow “hyperspectral” band spanning 400-to 2500-nm wavelength, has become increasingly feasible. In this technique, measurements are collected as three-dimensional data cubes with one spectral and two spatial dimensions. The spectra record the surface and atmosphere’s absorption and scattering properties, allowing material identification with overlapping but distinct spectral signatures.
Hyperspectral imaging is unique in its ability to resolve the limitations of conventional remote sensing techniques by making observations that cannot be achieved via other technologies. Nonetheless, there are still several obstacles to the widespread use of hyperspectral measurements.
Traditional satellite platforms (such as MODIS, Hyperion, VIIRS, AVHRR, and GOES), have prohibitively high costs to provide the required coverage and resolution, but the recent advances in small satellite technology allow for much shorter return times at a fraction of the cost of a traditional remote sensing mission. However, careful calibration and field testing of these new platforms is required before they can be utilized for a particular application. The HORIS platform is being applied to the remote sensing of wildland-urban interface fires and improve observational capabilities. Existing hazard mapping products leverage satellite observations to provide observational guidance, e.g., the NOAA Hazard Mapping System Fire and Smoke Product, the Landscape Fire and Resource Management Planning Tools, and USFS Wildfire Hazard Potential maps. However, in a changing climate, hazards evolve on shorter time and length scales than existing platforms can observe. The dynamic urban-wildland environment needs better observational coverage to constrain the ignition and evolution of fires.
The dataset delivered through a CubeSat LEO platform lends itself to assimilation into numerical and statistical models to improve reconstruction and forecasting of fire events. Specifically, the platform provides high-resolution measurements of:
• radiant heat via the onboard thermal IR sensor
• land albedo and fuel properties via the onboard visible imaging sensor
• vegetation index, burned area estimates and other land use properties via the onboard hyperspectral RedNIR sensor
• trends in such variables
This added capability will help improve the ability of models to constrain the efficiency of ignition onset, ignition sustenance, and burning of combustible material. The observational measurements will furthermore aid in the ability of models to constrain radiative and convective forces for major fire events.
The paper will outline advances in:
1. Onboard computing capabilities including low power alert camera
2. Scene-of-interest detection
3. AI detection of aerosol classes
4. Duty cycling
5. AI Scene Classifier performance
6. Airborne Payload performance and testing.
References:
Filchev, 2014: Satellite Hyperspectral Earth Observation Missions – A Review. Aerospace Research in Bulgaria. 26.
Jablonski et al., 2016: Best practices in passive remote sensing VNIR hyperspectral system hardware calibrations. SPIE Commercial Scientific Sensing and Imaging. 986004.
Gallagher and St Germain, 2018: The NOAA Satellite Observing System Architecture (NSOSA) Study, vol. 2018.
Anthes and Schreiner, 2019: Six new satellites watch the atmosphere over Earth’s equator, EOS Transactions American Geophysical Union, 100.
Dillon, 2015: Wildfire Hazard Potential (WHP) for the conterminous United States (270-m GRID), version 2014 classified. Forest Service Research Data Archive.
This work was funded under the U.S. National Oceanic and Atmospheric Administration Small Business Innovation Research program NOAA SBIR Award # NA23OAR0210337-T1-01.